Russian disinformation tactics via social media have been very topical in Western countries over the past year, and one of the specific tactics that has been spotlighted is the deceptive use of ‘bots’, or automated accounts, by Russia’s Internet Research Agency, particularly on Twitter. In a sense, bots hide in plain sight, purporting to be something that they are not. A useful function is therefore served if we can use data analytics techniques to automate the process of exposing bot activity in a way which helps direct an analyst’s attention towards the signature activity of bots within crowds of ‘genuine’ users. However, the problem is non-trivial because adversaries may deliberately introduce obfuscation in the form of slight differences between bots’ posts. Notwithstanding this, we show how the problem may be solved using the tensor decomposition method PARAFAC.